Compensated simulated annealing vs. dynamic programming used for boundary detection in intracoronary ultrasound

Abstract

Intracoronary ultrasound (ICUS) is a valuable tool in the study of coronary artery disease. Accurate and robust methods using real-time automated boundary detection algorithms are needed to quantify lumen and intimal-medial areas. The authors' approach transforms ICUS boundary detection into a recirculant multilayer graph problem with local minima. Dynamic programming (DP) and simulated annealing (SA) are two fundamentally different approaches to this optimization problem which are known to converge to the global minimum. However, time to convergence for SA is impractical. The authors compare a new optimized implementation of SA called compensated simulated annealing with DP.<>